Agree or Disagree: Predicting Judgments on Nuanced Assertions

摘要: Being able to predict whether people agree or disagree with an assertion (i.e. explicit, self-contained statement) has several applications ranging from predicting how many will like dislike a social media post classifying posts based on they are in accordance particular point of view. We formalize this as two NLP tasks: judgments (i) individuals and (ii) groups the text previous judgments. evaluate wide range approaches crowdsourced data set containing over 100,000 2,000 assertions. find that individual is hard task our best results only slightly exceeding majority baseline, but can be more reliably predicted using Siamese neural network, which outperforms all other by margin.